The nature of statistical learning theory
The nature of statistical learning theory
Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond
Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond
Credit rating analysis with support vector machines and neural networks: a market comparative study
Decision Support Systems - Special issue: Data mining for financial decision making
Methodology for long-term prediction of time series
Neurocomputing
Machine learning methods for microbial source tracking
Environmental Modelling & Software
System reliability forecasting by support vector machines with genetic algorithms
Mathematical and Computer Modelling: An International Journal
Hi-index | 0.03 |
The Sea Surface Temperature (SST) is one of the environmental indicators monitored by buoys of the Prediction and Research Moored Array in the Tropical Atlantic (PIRATA) Project. In this work, a year-ahead prediction procedure based on SST knowledge of previous periods is proposed and coupled with Support Vector Machines (SVMs). The proposed procedure is focused on seasonal and intraseasonal aspects of SST. Data from PIRATA buoys are used in various ways to feed the SVM models: with raw data, using information about the SST slopes and by means of SST curvatures. The influence of these data handling strategies over the predictive capacity of the proposed methodology is discussed. Additionally, the forecasts' accuracy is evaluated as the number of years considered in the SVM training phase increases. The raw data and the curvatures presented quite similar performances, they are more efficient than the slopes; the respective Mean Absolute Percentage Error (MAPE) values do not exceed 2% and all Mean Absolute Errors (MAEs) are lower than 0.37 ^oC. Besides, as the number of years considered in the training set increases, the MAPE and MAE values tend to stabilize.